高分辨率遥感影像道路提取方法研究

发布时间：2018-04-16 02:17

本文选题：高分辨率遥感影像 + 道路提取　；参考：《南京理工大学》2014年硕士论文

【摘要】：道路信息是一种重要的基础地理信息,从高分辨率遥感影像中提取道路可用于GIS数据的自动获取和更新、目标检测、图像匹配以及数字测图等,是人工智能、计算机视觉、遥感科学以及摄影测量等领域的研究重点之一,具有极其重要的科学意义与实用价值。本文对高分辨率遥感影像道路的提取方法进行了深入的研究,并结合高分辨率遥感影像中道路的特点,对现有的道路提取方法做出了一些有益的改进。首先,实现了基于GIS信息与Ribbon Snake模型的高分辨率遥感影像道路的自动提取。利用GIS数据构造的B样条曲线作为道路的初始中心线,并根据道路区域的纹理特征设计区域能量函数,通过自动调节各能量函数在总能量函数中所占的权重以及控制点的搜索范围,实现了模型的分级演化。实验证明了该方法在初始轮廓曲线远离目标位置情况下的有效性。其次,改进了path opening方法中的邻接图以提取曲率较大的道路,并将path opening与Ribbon Snake模型相结合,用于高分辨率遥感影像道路的自动精确提取。通过path opening操作获取初始道路中心线,由Ribbon Snake模型进行精确定位,得到准确光滑的道路中心线以及两条道路边界,通过实验验证了该方法的可行性与有效性。最后,提出了一种基于改进path opening的有损道路自动提取新方法。该方法巧妙地将path opening算法中的最大路径长度λ与区域的几何特征相结合,建立了道路区域的判定准则。通过实验对比分析,从实验效果和算法效率两个角度证明了该方法的有效性和优越性。
[Abstract]:Road information is an important basic geographic information. Road extraction from high-resolution remote sensing images can be used for automatic acquisition and updating of GIS data, target detection, image matching and digital mapping. It is artificial intelligence and computer vision.Remote sensing science and photogrammetry are of great scientific significance and practical value.In this paper, the road extraction methods in high-resolution remote sensing images are studied deeply, and some useful improvements are made to the existing road extraction methods combining with the characteristics of roads in high-resolution remote sensing images.Firstly, the road extraction of high resolution remote sensing image based on GIS information and Ribbon Snake model is realized.The B-spline curve constructed by GIS data is used as the initial centerline of the road, and the energy function of the region is designed according to the texture features of the road area.By automatically adjusting the weight of each energy function in the total energy function and the search range of the control point, the hierarchical evolution of the model is realized.The experimental results show that the method is effective when the initial contour is far from the target position.Secondly, the adjacent map in path opening method is improved to extract roads with large curvature, and the path opening and Ribbon Snake model are combined to extract the road automatically and accurately from high-resolution remote sensing images.The initial road centerline is obtained by path opening operation, and the accurate and smooth road centerline and two road boundaries are obtained by the Ribbon Snake model. The feasibility and effectiveness of the method are verified by experiments.Finally, a new method for automatic extraction of lossy roads based on improved path opening is proposed.The method skillfully combines the maximum path length 位 of path opening algorithm with the geometric characteristics of the region, and establishes the criterion of road region.The effectiveness and superiority of this method are proved from two aspects of experimental effect and algorithm efficiency.
【学位授予单位】：南京理工大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：U495;TP751

【参考文献】